Re: [tied] Demonstratives

>>> Hitt. we:s and Gothic we:s are from

*wey-es.

>>
>>Before accepting, it needs to be determinated

whether intervocaloic *-y- was lost in Hitt. and Goth. I don't know about Gothic, but Hittite doesn't seem to have such a soundlaw.

> Yes it

has. Melchert, p. 130: "Intervocalic */y/ is
regularly lost in Hittite".

Well, Melchert is an authority, and we must trust him. But in declension of i-stems (nom. -is vs. gen. -ayas) the *-y- has not been lost!

> I see that Starostin reconstructs them for Proto-NW
>

Caucasian (Abxaz-Adyghe). Like PIE, NW Caucasian has also

> lost the

vowels *i and *u, and the vocalic features of

> labialization and

palatalization have been transferred en

> masse to the consonants.

Something similar no doubt

> happened in PIE.

> According to

L&M, labialized labials also occur in living

> languages of the

Caucasus (and elsewhere), but I wouldn't

> know off-hand which

ones.

In this light the possibility of pre-PIE *mW is more acceptible.

> In any case, there's nothing impossible about labialized
>

labials. They're just more likely to be lost quickly than

>

labialized dentals, which are more likely to be lost than

> labialized

velars. PIE has only retained the labialized

> velars *kW, *gW and

*ghW as full-fledged phonemes. There

> are traces of *sW and

*tW. The labialized labials can only

> be deduced from irregular

sound correspondences within IE

> (*m ~ *w [1pl. *-men ~ *-wen], *p ~ *kW

[*wl.p- ~ *wl.kW-

> "wolf"], *bh ~ *ghW [*bhen- ~ *ghWen-

"beat"], etc.).

To reconstruct a distinct phoneme, it is necessary to demonstrate that its introduction allows to see new regularities in the case where only irregularities were seen before. Reconstructing labialized labials does not allow to understand more clear the correspondences of *wl.p- ~ *wl.kW- etc. Unfortunately, I don't know in what languages the reflexes of *wl.p- exist, so I'll take another example: *akW- ~ *ap- "water". We see *akW- in Latin, but *ap- in Sanscrit. If here the reflexes of *pW are found, the following soundlaw has to exist: *pW > Lat. kW, Scr. p. But if so, we must expect *wl.pW > Scr. vr.p-, but really Sanscrit has vr.k-. So, such a reconstruction gives nothing here.

Further, on 1pl. *-men ~ *-wen. As long as I know, this alternation is found only in Anatolian, and therefore there exists a possibility that it's only an Anatolian innovation in phonetical reflexes of PIE *-me-. But 1du. *-we (if survives) exists everywhere (e.g. Old Slavic esve "we two are" vs. esmy "we many are"). Can you explain it from *-mWe?

> Hittite 2sg. -ti is the hi-conjugation ending, presumably
> from

*-th2a-i. Even if it were from *-th2-i directly, that

> does not

consitute evidence against the solid soundlaw *ti >

> (z)zi: the

cluster *th2 might simply have merged with *dh

> here, and the regular

outcome is *dhi > ti (and *di > si).

Hittite 2sg. -ti is used also with some verbs of mi-conjugation, and as it is obligatory with them, it is not possible to see here only the influence of the hi-paradigm.

>>In fact, the question on Hittite palatalization *ti > zi

is not so clear as it is commonly assumed. Sometimes it no doubt exists, but there are examples where the palatalization is not found.

> Such

as?

Now I remember nothing but the mentioned -ti < *-ti, but such examples exist. Later I'll try to cite them.

> I don't see the relevance of English 3sg. -s.

Why? It is a very misterious feature, and germanists fail to explain from where it is. But if we assume that it is an IE archaism that was preserved in some dialects, and further introduced in litterary English in the place of -th... Really, my explanation is very attractive. :)

> German 2sg. -st is 2sg. *-s plus the 2nd. person pronoun

*þu,

> du (e.g. bis du > bistu > bist du). Tocharian does have

> undeniable 2sg. ending *-t, but it's not necessary to derive
>

that all the way back to PIE: it is most likely either a

> middle/stative

ending *-th2a(-), or an agglutinated personal

> pronoun *tu.

Such explanations are, of course, possible as well... But you need to invent a separate hypothesis for each case, while all of them are easily solved if we accept that PIE had in 2Sg. both *-s and *-t.

>>It is a very strange situation when certain sounds are more

used in suffixes than in roots.

> It's rather common, actually.

To give just one example: in

> PIE, the long diphthong *-o:i is very

common in grammatical

> endings (dat.sg., ins.pl.), but very rare

elsewhere.

I don't think that *o:i here was a single phoneme. It would be better to interprete it as *o: + *i/*y (or even *o + *o + *i/*y). So the example is not valide.